I.Zaytsev's profile - activity

Hello! I have several of the following images:

image1

image2

image3

I have a task to recognize the text on the labels. I will recognize the text using the FineReader Engine (special SDK for recognize images). However, I need to do preprocessing of images for better recognition. What do I need to do to remove barrel distortion and other distortions?

For example, FineReader bad recognizes the "BordeauX" word in the first image(image1), it instead of the "BordeauX" word sees "ordeauX", may be because there is a flare? How can this be fixed?

Have any idea how to do the preprocessing of the images?

Hi, all!
I want to detect largest rectangle. This source image.
I write this code:

def rectangle_detection():
    src = cv2.imread("rectangles.jpg")
    if src == None:
        print "Do not open image!"
        return -1

    gray = cv2.cvtColor(src, cv2.cv.CV_BGR2GRAY)
    blur = cv2.blur(src, (3,3))
    canny = cv2.Canny(gray, 100, 100, 5)
    lines = cv2.HoughLinesP(canny, 1, cv2.cv.CV_PI/180, 70, 30, 10)

image after HougLinesP()
how to draw the lines int the sorce image? and what should I do next?
Thanks!

ok! thanks :)

OpenCV are similar for different programming languages
Compare them and you find solution...

this code for draw line around pupil(python):

   gray = cv2.cvtColor(~src, cv2.cv.CV_BGR2GRAY)
    cv2.threshold(gray, 220, 255, cv2.THRESH_BINARY, gray)

    contours, hierarchy = cv2.findContours(gray.copy(), cv2.RETR_EXTERNAL, cv2.CHAIN_APPROX_NONE)

    cv2.drawContours(gray, contours, -1, cv2.cv.RGB(255,255,255), -1)

    i = 0
    while i < len(contours):
        area = cv2.contourArea(contours[i])
        rect = cv2.boundingRect(contours[i])
        radius = rect[2] / 2 #rect[2] - width
        i = i + 1
        if area >= 30:
            if abs(1 - (rect[2] / rect[3])) <= 0.2: #rect[3] - height
                if abs(1 - (area / (cv2.cv.CV_PI * pow(radius, 2))) < 0.2):
                    cv2.circle(src, (rect[0] + radius, rect[1] + radius), radius, cv2.cv.RGB(255, 0, 0), 2)
                    cv2.imwrite(str("") + str("1") + ".jpg", src)
                    break

I think that you will have a similar situation

thanks! it's work:
contours, hierarchy = cv2.findContours(gray.copy(), cv2.RETR_EXTERNAL, cv2.CHAIN_APPROX_NONE)
cv2.drawContours(gray, contours, -1, cv2.cv.RGB(255,255,255), -1)

HellO! I am russian student and I began to study OpenCV :) I write this code:

import cv2
import numpy

def eye_detection():
    src = cv2.imread("E:\\Ilya\\PythonProjects\\eye.jpg")
    if src == None:
        print 'Do not open image!'
        return -1

    gray = cv2.cvtColor(~src, cv2.cv.CV_BGR2GRAY)
    cv2.threshold(gray, 220, 255, cv2.THRESH_BINARY, gray)

    contours = cv2.findContours(gray.copy(), cv2.RETR_EXTERNAL, cv2.CHAIN_APPROX_NONE)
    #cv2.imwrite(str("E:\\Ilya\\PythonProjects\\") + str("1") + ".jpg", gray)
    #print(contours)
    numpy.array(contours, dtype=numpy.float)
    cv2.drawContours(gray, cv2.cv.Scalar(contours), -1, cv2.cv.RGB(255,255,255), -1)
if __name__ == "__main__":
    eye_detection()

If I compile this script:

line 16, in eye_detection
    numpy.array(contours, dtype=numpy.float)
ValueError: setting an array element with a sequence.

contours contains(print(contours)):

([array([[[344, 254]]]), array([[[194, 243]],

       [[194, 244]],

       [[195, 244]],

       [[196, 244]],

       [[197, 244]],

       [[198, 244]],

       [[199, 244]],

       [[200, 244]],

       [[201, 244]],

       [[202, 244]],

       [[203, 244]],

       [[204, 244]],

       [[205, 244]],

       [[204, 244]],

       [[203, 244]],

       [[202, 244]],

       [[201, 243]],

       [[200, 243]],

       [[199, 243]],

       [[198, 243]],

       [[197, 243]],

       [[196, 243]],

       [[195, 243]]]), array([[[217, 242]], etc.

how to solve this problem?